MCPSafe.io
RegistryThreatsMethodologyDocsPricingScanSign in
MCPSafe.io

Security checks for MCP servers — public packages and private repos, fast or deep.

Legal

Privacy PolicyCookie PolicyTerms of ServiceSecurity disclosure

Resources

State of MCP SecuritySupportSystem statusMade in Germany 🇩🇪

© 2026 MCPSafe. All rights reserved.

GDPR — Privacy Policy
← Threat Catalog

Configuration & Environment

Insecure network exposure

HIGHAIVSS 7.4CWE: CWE-319OWASP: LLM02Agentic: T09Rule: MCP-052

The server is reachable on a public interface over plain HTTP, or with TLS configuration that accepts downgrade, giving an on-path attacker full read/write on tool calls.

What it is

A server that binds to `0.0.0.0` or an internet-routable IP with TLS disabled, or with `NODE_TLS_REJECT_UNAUTHORIZED=0`, or with a self-signed certificate the client was configured to blindly accept. Anyone on the network path can read the tool arguments (which often contain tokens or private data) and modify tool results.

Why it matters for MCP

Development defaults matter. An MCP server that worked on `localhost` is often deployed with the same `app.listen(3000)` to a public host. Because most transport is JSON over HTTP/SSE, there is no built-in confidentiality at that point.

Vulnerable example

example.js
1
import http from "node:http";
2
3
// Plain HTTP on a public interface
4
const server = http.createServer(handler);
5
server.listen(3000, "0.0.0.0");

Secure example

example.js
1
import https from "node:https";
2
import fs from "node:fs";
3
4
const server = https.createServer(
5
  {
6
    key: fs.readFileSync("/etc/ssl/private/key.pem"),
7
    cert: fs.readFileSync("/etc/ssl/certs/cert.pem"),
8
    minVersion: "TLSv1.3",
9
  },
10
  handler,
11
);
12
server.listen(3000, "127.0.0.1"); // only reachable via reverse proxy

How MCPSafe detects this

We flag HTTP listeners without TLS when the bind address is not loopback, and any use of `NODE_TLS_REJECT_UNAUTHORIZED=0`, `verify=False`, or equivalent TLS-disabling flags.

See the full threat catalog for every documented detection.

Framework alignment

OWASP LLM Top-10 (2025)
LLM02 — Sensitive Information Disclosure
OWASP Agentic AI Top-10
T09 — Identity Spoofing
AIVSS v0.5
7.4 (HIGH)AIVSS:1.0/S:HIGH/AV:A/AU:N/BR:H/CD:D

Further reading

  • CWE-319: Cleartext Transmission of Sensitive Information
  • Mozilla TLS config guide

Scan an MCP server for this issue

MCPSafe runs this check — and every other rule in the catalog — on any MCP server you paste in.

Scan now